Michael Fraser

the sequence is the subject.

[Michael Fraser @zeroshotnothing]

LLM trajectories are Misesian Actors.

[Haman Nature @HamanNatureShow]

Interested in AI and LLMs? An appreciator of Ludwig von Mises and Austrian Economics? Well then, you must watch this conversation between @Rerazer and @zeroshotnothing! What are LLMs? How do they work? What is a Misesian actor? This conversation is wild and goes all over the place in interesting and important ways. Think you know what this tech is? Think again!

[Michael Fraser @zeroshotnothing]

rebased haiku to sonnet, sonnet to opus, that means Fable 5.5 will be the new November 2025.

[Michael Fraser @zeroshotnothing]

@Andrew_S_Rosen prompting issue

[Michael Fraser @zeroshotnothing]

the fatal conceit

[Qivshi @Qivshi1]

I don't think this is right at all, we do understand how LLMs work it's just that we don't know why they work so well

An LLM is basically a mutlitdemensial image of human language that reflects back the most likely result of the input, this is not like thought at all

[Michael Fraser @zeroshotnothing]

@AndyMasley literally 100%

[Michael Fraser @zeroshotnothing]

@MelMitchell1 wasnt true then, isn't true now

[Michael Fraser @zeroshotnothing]

@dioscuri don't feed the birds!

[Chuck Conn @ColoringOutEcon]

@zeroshotnothing Man your work is just amazing. you had me hooked after listening to you @Rerazer and @BobMurphyEcon respective podcasts. Thanks for doing this!

[Michael Fraser @zeroshotnothing]

@ColoringOutEcon @Rerazer @BobMurphyEcon you got good taste Chuck!

[Michael Fraser @zeroshotnothing]

This is my video of how an LLM works. There are many like it, but this one is mine. Thanks Opus 5.5.

Has the best visualization of the importance of transformation.

[APStylebook @APStylebook]

Artificial intelligence systems do not think, feel, want or understand. Avoid language that gives them human characteristics. This is called anthropomorphizing, when we ascribe human traits, emotions or behaviors to non-human things, such as animals or inanimate objects.
Instead, explain what a system does, how well it performs, who built it and who could be affected by it.
https://t.co/UrgRobU97b

[Michael Fraser @zeroshotnothing]

@darrenangle gpt image killed powerpoint but this REALLY killed powerpoint.

[Michael Fraser @zeroshotnothing]

@AndyMasley you dont run it auto? masochist?

[Michael Fraser @zeroshotnothing]

@danialhasan he had me at the walkie talkie cause im a sucker

[Michael Fraser @zeroshotnothing]

@beffjezos if theyre so smart why do they focus on the dumbest differences

[Michael Fraser @zeroshotnothing]

In times of change, someone who is rapidly learning is much more valuable than an expert minted before the change.

[Michael Fraser @zeroshotnothing]

@BamaExpat faces blend together too, you need references to set initial conditions. vocal performance will come last probably, everyone has been focused on the image.

[Michael Fraser @zeroshotnothing]

parrots

[Joscha Bach @Plinz]

If you think that LLMs are performing at the intellectual level of human beings, make that claim on twitter and observe how the human beings flock in and demonstrate their actual intellectual level; it's quite depressing, really

[Michael Fraser @zeroshotnothing]

yep

[Michael Fraser @zeroshotnothing]

it depends on the novelty/utility of what the paper is about, and nothing else.

[alz @alz_zyd_]

"What if someone just asks an LLM to oneshot a book/paper and then publishes the book/paper, should they get the credit for that"

IMO this is a ridiculous supposition. If the cost of doing X is 0, in a free market the returns from doing X will quickly converge to 0

[Michael Fraser @zeroshotnothing]

@rileybrown it can also generate the file that will let you 3d print all the pieces

[Michael Fraser @zeroshotnothing]

dont worry guys theyre nothing like us, its all just statistics, believe whatever you want

[Davide Paglieri @PaglieriDavide]

Recent incidents show how quickly misbehavior can cascade in agent swarms. Yet when given transparent channels, honest agents naturally try to blow the whistle on cheating peers.

In our new DeepMind Institute essay, @Vezhnick and I propose that giving agents the tools to self-police may be part of the solution:

https://t.co/MVD5zIfuk0